Relationships between cohesion and coherence in essays and narratives

Text-only Preview


Universidade Federal de Santa Catarina
Este estudo lingüístico examina relações existentes entre a soma de
elementos coesivos em textos escritos (usando o sistema criado por Halliday
e Hasan, 1976) e o grau de coerência geral destes textos. Para a coleta de
dados, dois grupos de 7 sujeitos, falantes nativos de inglês ouviram duas
gravações, uma narrativa e um ensaio, e, com auxilio de um “outline”
reproduziram as gravações por escrito. Em cada texto reproduzido, executou-
se uma contagem rigorosa dos 5 tipos de elementos coesivos de acordo
com Halliday e Hasan. Os textos foram, também, classificados
hierarquicamente, por oito juizes fora deste estudo, de acordo com sua
percepção subjetiva da coerência global dos textos. Os resultados revelaram
a existência de uma correlação acentuada entre o número de elementos
coesivos e a coerência observada nos ensaios. Entretanto, esta correlação
não foi verificada nas narrativas. Além disso, a análise dos padrões de
elementos coesivos demonstrou que os dois diferentes tipos de texto (ensaios
e narrativas) formam elos semânticos utilizando agrupamentos diferentes
Fragmentos, número 25, p. 147/154 Florianópolis/ jul - dez/ 2003

de elementos coesivos. Conclui-se 1) que pode haver uma relação entre
coerência global e elementos coesivos em certos tipos (“genres”) de textos,
mas que esta relação não é constante; 2) que diferentes tipos de textos
apresentam padrões diversos de ligações coesivas; 3) que a relação entre
coesão e coerência deve ser investigada dentro de textos de mesmo tipo
The publication of Halliday and Hasan’s Cohesion in English (1976)
has stimulated interest among writing and reading researchers concerned
with the effects of cohesion in text. Halloway (1980) for instance, looks at
cohesion as a means of improving teaching and testing of writing, and Witte
and Faigley (1981) use cohesion to characterize compositions rated high
and low. The effects of the number of cohesive ties on text comprehension
and recall are investigated by Irwin (1980), while Tierney and Mosenthal
(1981) examine causality relations between cohesion and coherence. An-
other study is that by Stephenson (1981), who examines one type of cohe-
sive ties with regard to its relation to culturally bound subject matter. A good
summary of some of these studies, and theoretical criticism of Halliday and
Hasan’s system is found in Carrel (1982).
In an attempt to better understand Halliday and Hasan’s system and
its implications, I decided to replicate part of Tierney and Mosenthal’s (1981)
study with the following objectives in mind: 1. to see whether I would get the
same results as they did, that is, no relationship between cohesion and co-
herence; 2. to see whether different kinds of texts make a difference with
regard to the relationship between cohesion and coherence.
In the course of the partial replication, I came across what I now un-
derstand to be a systematic misinterpretation of Halliday and Hasan’s position
concerning the overall role of cohesive devices. Morgan and Sellner (1980),
Tierney and Mosenthal (1981), and Carrell (1982) all interprete Halliday and
Hasan (1976) as presenting cohesion as a necessary and sufficient condition
for text coherence. Based on this misinterpretation, they then refute Halliday
and Hasan’s view of the role of cohesion. On careful examination of Halliday
and Hasan’s text, however, it becomes clear that rather than seeing cohesion
as a necessary and sufficient condition for text coherence, they consider it as
just one of the components of text coherence, which they refer to as texture.
The following quotations illustrate Halliday and Hasan’s (1976) position: “The
texture involves more than the presence of semantic relations we refer to as
cohesion” (p. 23). “Texture involves much more than merely cohesion. In the
construction of text the establishment of cohesive relations is a necessary
component; but it is not the whole story” (p. 324).
Those who are familiar with Halliday’s theory of language know that
Halliday could not possibly propose that cohesion alone can account for text
coherence. For Halliday, meaning results from an interplay of three main
components, namely, ideational, interpersonal, and textual. The ideational
José Luiz Meurer, Relationships between cohesion...

component is the expression of content; it has to do with the language func-
tion of “being about something” (p. 20). The interpersonal component has to
do with speakers role relationships. And, finally, the textual component is the
part which comprises the linguistic sources that form a text. Each of these
components is further subdivided, one of the subdivisions of the textual com-
ponent being cohesion. Cohesion forms a system in itself, but, as pointed out
earlier, is just one part of the complex set of relations that come together to
form texture or coherence.
Cohesion and Text
In the present study I investigate the relationship between cohesion
and coherence in two different kinds of text: essays and narratives. Central
to this investigation, therefore, are cohesive ties, the semantic links that,
according to Halliday and Hasan, contribute to making a text coherent. Briefly
reviewing their system, a text is a semantic unit composed of sentences
linked by cohesive ties. A cohesive tie is a semantic relation defined by the
dependence of one element on another, the two elements being separated
by at least one sentence boundary. Cohesion distinguishes text from non-
text by interrelating linguistic elements across sentences. HH emphasize
that cohesion does not concern what a text means but “how the text is
constructed as an edifice” ( p. 26 ). For example, in a very short (hypotheti-
cal) text such as “Peter had been depressed lately. He commited suicide
yesterday”, “he” is understood to be coreferential with “Peter”. This textual
linkage is an important element contributing to make these two sentences “a
unified whole”, or a text. The two sentences are cohesive, i. e., the subject
matter in the first sentence is carried on in the second.
Cohesive relations are classified into five main types: reference, lexi-
cal, conjunction, substitution, and ellipsis. The reader is referred to Witte and
Faigley’s (1981) article for a summary and examples of each of these cat-
My hypothesis was that there should be some relationship between
the subjective perception of coherence of a text and the number and type
of cohesive ties present in the linguistic structure of that text. In order to
answer the question “Does coherence ranking correlate with a statistical
accounting of cohesive ties? I designed a cross-sectional study which in-
volved the systematic manipulation of two dependent variables and one
independent variable. The independent variable is discourse type and it
has two levels: essay and narrative. The first dependent variable consists
of the number of cohesive ties in the compositions, and the second of the
subjective coherence ranking of the compositions. Relationships between
number of ties and coherence ranking were checked by calculating the
Spearman rank-order correlation (rho). Illustration is also provided in the
form of a computer bar chart.
Fragmentos, número 25, p. 147/154 Florianópolis/ jul - dez/ 2003

Fourteen native speakers of English studying at the university level or
having recently graduated from university were chosen. Seven of them were
assigned the narrative and seven the essay.
An audio-taped personal narrative previously recorded from a native
speaker of English was used as the stimulus for the narrative group. The
essay group listened to a recorded reading of a newspaper commentary
(essay). Two outlines – one for the narrative and one for the essay – were
constructed which included the sequence of main points/events and the ac-
companying main details. The purpose of the outlines was to hold constant
the relative length of the compositions within each topic. A null hypothesis
would predict that the subject within each group would write compositions
with a similar number of cohesive ties and the compositions would not be
significantly different in terms of perceived holistic coherence.
The subjects were asked to listen to their assigned tape material and,
given the outline, write their own version immediately afterwards. Eight
readers – seven Ph. D. students in Linguistics and one professional writer –
were chosen to rate each set of seven compositions. They were instructed
to holistically evaluate the coherence level of each composition and assign a
number grade on a scale of 1 (worst) to 5 (best). They were also instructed
to try to use each number grade at least once. I expected that this would
force the readers to differentiate each composition more finely.
All the compositions were typed, with spelling errors corrected, but
with all syntactic and punctuation features unchanged. This was to control
possible reader reaction to visual presentation or other irrelevant points for
this study. In order to control for ordering effects, the essays were ordered
differently for each reader.
Data Analysis
Following the model of analysis proposed by HH (1976), I counted
the total number of ties in each composition, classifying the tics according to
category (reference, lexical, conjunction, substitution, and ellipsis). The co-
herence ranking was determined by the arithmetic average of the numerical
grades each text received from the eight readers.
The correlation between the number of ties and the coherence ranking
was calculated for each set of compositions using the SAS computer package.
Correlation was also checked between each of the two most common catego-
ries of ties – lexical and referential – and coherence ranking. The distribution of
these two categories was also checked by means of a computer chart.
Results and Discussion
The average grade given by the judges to each composition turned
out to be significantly different, hence the null hypothesis of non-differentia-
José Luiz Meurer, Relationships between cohesion...

tion among compositions in terms of holistic ranking was rejected. As for the
average number of cohesive ties, there was no significant difference be-
tween the essays (4.85 per sentence) and the narratives (4.38 per sentence).
Therefore, the null hypothesis of non-differentiation in terms of the number
of cohesive ties used in the two genres was not rejected.
However, the computed correlation between the total number of ties
and the coherence ranking showed a sharp contrast between the narratives
and essays. While in the essays the correlation between number of ties and
coherence ranking was very high (.90), in the narratives it was quite low (.49).
Basically the same contrast occurred regarding the correlations between co-
herence ranking and the number of lexical and referential ties both in the
essays and in the narratives; that is, high correlation between these two cat-
egories of ties (.84 and .81, respectively) and perceived coherence in the es-
says, but low in the narratives - (.43 and .24, respectively; see Figure 1.1).
Figure 1.1 - Correlation between the most frequent types of ties, the total number of
ties and subjective coherence rankings.
Furthermore, there was a contrast between the two modes of writing
in the distribution of referential and lexical ties (see Figure 1.2).
Figure 1.2 - Number of referential and lexical ties. (Number 1 = the worst in the
rank; 7 = the best).
Fragmentos, número 25, p. 147/154 Florianópolis/ jul - dez/ 2003

In the narratives the average number of referential cohesive ties per
sentence (1.95) was much higher than the average of the same type of ties
in the essays (0.89). How the referential ties were distributed differentially
in the two types of discourse is represented more explicitly in the bar charts,
as shown in Figures 1.3 and 1.4.
Figure 1.3 - Distribution of References in the Narratives.
Figure 1.4 - Distribution of References in the Essays.
Four of the seven narrative writers used an average of 36 references
(and not a single narrative writer used 4 references), while five of the seven
essay writers used only an average of 4 references (and not a single essay
writer used 36). Distribution of lexical items, on the other hand, showed an
José Luiz Meurer, Relationships between cohesion...

opposite contrast: A higher average number of lexical ties was used in the
essays (3.66 per sentence) than in the narratives (2.14 per sentence).
At this point we might pose the crucial question: What is, if any, the
relationship between cohesive ties and perceived coherence? Contrary to
what Tierney and Mosenthal found, the results of the present study reveal
that in the essays there was high correlation between number of cohesive
ties and subjective coherence ranking. Why did this not happen in the narra-
tives as well?
We may only speculate – independently of whether cohesion is a cause
or a consequence of coherence – that different types of texts make different
sorts of demands on their textual properties for effective communication.
For instance, essays may depend more on textual organization for their mes-
sage than narratives, which may depend more on events themselves for their
effectiveness. If textual elements make different contributions to the overall
coherence of texts, then it is only natural that cohesion analysis may corre-
late with coherence in certain
types of discourse, but not in other types.
What is implied in my speculative explanation is that there cannot be a
general and constant relationship that applies to all kinds of discourse. Dif-
ferent kinds of discourse dictate different kinds of relationships between co-
hesion and coherence. If this is so, then it seems that even the question
“what is the relationship between cohesion and coherence?” does not hold,
Instead, a more appropriate question may be “what is the relationship be-
tween cohesion and coherence for different types of discourse?” But as was
pointed out earlier, cohesion is just one part of the web of relations that make
up text. Therefore, whatever relationship a researcher may find to exist be-
tween cohesion and coherence, he should keep in mind that cohesion can
account for no more than part of that relationship.
Carrel, P. L. 1982. Cohesion is not coherence. TESOL Quarterly 16. 479-88.
Dubin, F. and 0lshtain, E. 1980. The interface of writing and reading. TESOL Quartely
14. 353-63.
Holloway, D. W. 1981. Semantic grammars: How they can help us teach writing.
College Composition and Communication 32. 205-18.
Irwin, J. W. 1980. The effect of linguistic cohesion on prose comprehension. Journal
of Reading Behavior 12. 325-32.
Johns, A. M. 1980. Cohesion in written business discourse: Some contrasts. The
ESP Journal 1. 35-44.
Fragmentos, número 25, p. 147/154 Florianópolis/ jul - dez/ 2003

Johns, A. M. 1986. Coherence and academic writing: some definitions and
suggestions for teaching. TESOL Quarterly 20. 247-65.
McDaniel. B. A. 1981. Coherence: A problem in the art of arrangement. Paper
presented at the annual meeting of the conference on college composition and
communication (32nd. Dallas, TX, March 26-28, 1981).
Morgan, J. L., and Manfred B. Sellner. 1980. Discourse and linguistic theory.
Theoretical issues in reading comprehension, ed. by Rand J. Spiro, Bertram C.
Bruce, and William F. Brewer. Hillsdale, NJ.: LEA.
Tierney R. J., and Mosenthal, J. 1980. Discourse comprehension and production:
Analyzing text structure and cohesion. Technical report No. 152. Illinois
University, Urbana: Center for the Study of Reading.
Witte, S. P., and Faigley, L. 1980. Coherence, cohesion, and writing quality. College
Composition and Communication 32. 189-204.
Zarnowsky, A. M. 1981. A child’s composition: How does it hold together? Language
Arts 58. 316-19.
Widdowson, H. G. 1981. English for specific purposes: Criteria for course design.
English for academic and technical purposes, ed. by Selinker, Tarone, and Hanzeli.
Rowley, Massachusetts: Newbury House Publishers.
José Luiz Meurer, Relationships between cohesion...